Unit validation with distrib uniform not working

[matthiaskoenig]

Hi all,
I have an SBML model with an initial assignment from a uniform distribution (model attached).
spt_random.zip
Despite having set the units on the distributions I get a unit validation warning.
It seems like the unit validation does not work with the distributions.

Here the basic structure. An initial assignment from a uniform(0 dimensionless, 1 dimensionless) to a parameter which is dimensionless. I.e. the distribution and the parameter are compatible, but still get the warning below.

<listOfParameters>
  <parameter id="protein_random" value="0.5" units="dimensionless" constant="true"/>
</listOfParameters>
<listOfInitialAssignments>
  <initialAssignment symbol="protein_random">
    <math xmlns="http://www.w3.org/1998/Math/MathML" xmlns:sbml="http://www.sbml.org/sbml/level3/version2/core">
      <apply>
        <csymbol encoding="text" definitionURL="http://www.sbml.org/sbml/symbols/distrib/uniform"> uniform </csymbol>
        <cn sbml:units="dimensionless" type="integer"> 0 </cn>
        <cn sbml:units="dimensionless" type="integer"> 1 </cn>
      </apply>
    </math>
  </initialAssignment>
</listOfInitialAssignments>

WARNING  E0: SBML unit consistency (core, L91, code)                                                                                                                                                          validation.py:186
         [Warning] Missing unit declarations on parameters or literal numbers in expression                                                                                                                                    
         In situations where a mathematical expression contains literal numbers or parameters whose units have not been declared, it is not possible to verify accurately the consistency of the units in the                  
         expression.                                                                                                                                                                                                           
          The units of the <initialAssignment> <math> expression 'uniform(0, 1)' cannot be fully checked. Unit consistency reported as either no errors or further unit errors related to this object may not                  
         be accurate.      

Probably similar things happen with the other csymbol distributions.

Best Matthias

[fbergmann]

looks like unit validation cannot know what the result of a function defined in the distrib package is. I am not aware of any plugin system in libSBML that allows packages to extend unit validation. Will have to discuss with @skeating

[matthiaskoenig]

Thanks.
I tried the following 'uniform(0, 1) dimensionless', i.e. annotating the unit of the csymbol similar to constants in math expressions (e.g. 1 dimensionless`). Not sure if something like this could work to communicate the unit.

[fbergmann]

it feels like the proper thing to do would be to have the distrib plugin provide information about the csymbols and the unit input/output relationship. adding dimensionless like in your example would prove troublesome in more involved expressions, potentially.

[matthiaskoenig]

Hi all,
is there some simple solution to support unit validation for the distributions? I have some models using the distrib features with units and these errors make it hard to unit validate the models.
Best Matthias

[matthiaskoenig]

Thanks so much.

[fbergmann]

we'll try to get new binaries out next week